Selected papers by Dr. W. Edwards Deming
Dr. Deming published over 170 articles, wrote numerous unpublished papers for his students and clients, and conducted hundreds of studies for clients. These and numerous other writings by Dr. Deming are in the National Archives, The Library of Congress (LOC) in Washington, DC. To access this collection, call the LOC Manuscript Division. Since access to the collection is restricted, please call the LOC at 202-707-5387 to receive an access form.
Click on the icon to read the synopsis. icon will open the article in a new tab.
Articles are sorted by title. Click here to sort by year
Authored by Dr. Deming and T. Nelson Grice, this article illustrates how stratified sampling and ratio-estimates may increase the efficiency of a sampling procedure under appropriate conditions. They include the sample design, sampling procedure, why direct confirmation was not practicable, other auditing procedures possible, advantages of the statistical method, and gains from the ratio estimate.
This six page article is reprinted from Management Accounting, March 1970.
A study will be conducted to estimate the overall proportion of people that are affected with some defined psychopathology. The final determination of the psychiatric and other medical characteristics of a person will be made by a psychiatrist. A plan to use the services of trained interviewers to screen and separate into two classes (with and without apparent psychopathology) a large preliminary sample in order to conserve the time of the psychiatrist, by letting him test mainly cases that are almost surely afflicted with psychopathology, is appealing wherever the cost per case is much lower for the screening than for the psychiatric examination. It is not generally appreciated, however, that the screening-test, to be economical, must be relatively cheap and must admit only a low proportion of false negatives. This principle is not new, but illustrative calculations that show how false negatives affect costs, and why false positives are not so important, are hard to find in the literature. The purpose here is to present some theory and a simple illustration.
This eight-page article was prompted by Dr. Deming's work on a number of studies that required use and extension of the theory presented here. It was published in International Statistical Review, 45 (1977) 29-37.
Dr. Deming, as a practicing statistician, wanted to make clear the relationship between consultant and client, so there would be no questions about who was responsible for what. After discussion of the aim of his client engagements, Dr. Deming details the obligations of the client, and the obligations of the statistician to the client. Additional direction is given about interpretation of the results, and the statistical report or testimony related thereto. A section on fees and another on proprietary rights in statistical procedures are also included.
"Code of Professional Conduct" appeared in the International Statistical Review, vol.40, Aug. 1972: pp.215-219. It is a revision of a paper by the same title that he had written six years earlier.
Dr. Deming wrote this article to comment on recommendations made by an auditing standard of the American Institute of Certified Public Accountants, to explain why those recommendations conflicted with statistical practice. In addition to identification of indefensible methods in auditing, Dr. Deming provides a solid primer on Cause-System, Frame, Random Sampling, Stable Cause-System in Manufacturing, and the Likelihood Ratio.
The twelve-page article appeared in the Journal of Accounting, Auditing & Finance, Vol 2, No.3, Spring 1979, pp. 197-208.
When this paper was written, the cost of handling freight on the platforms of motor carriers of general freight amounted to easily 20% of the total cost of transportation of general freight by truck. In this paper, Dr. Deming aimed to learn how certain costs on the platform vary with the weights of shipments. He studied the total minutes required to handle freight on the platform by method and movement, for selected weights from 50 pounds to 15,000 pounds . Time was recorded separately for "truck to platform", "platform to truck", and "truck to truck" movements. Categories analyzed were (1) manual handling, no dragline or dragline not used, (2) manual handling at terminals where dragline is used, and (3) fork lift. Although data used were from the "Motor Carrier Platform Study", June 1973, Interstate Commerce Commission, the same methods of analysis can be used today. Dr. Deming includes a summary of results, source of data, method of analysis, and estimates of parameters and their standard errors.
This nine-page article appeared in Transportation Journal, Vol 17, No.4, 1978, pp. 5-13.
Of every dollar spent for transporting articles moved as common general freight by motor carrier in the 1970s, 13 cents was for stop-time at pickup and delivery. In this paper Dr. Deming explored the relationship between stop-time and weight. Using Interstate Commerce Commission cost data Dr. Deming fit a least squares line to the data for each region, calculated coefficients and standard errors, and developed the mathematical relationship between weight and man-minutes required to handle shipments. He then compared his results with the Commission's procedure for the computation of minutes per shipment.
This seven-page article was published in the Transportation Journal, Vol. 18, No. 2, 1979: pp.79-85.
Dr. Deming's aim in this paper is to contribute something to the improvement of statistical practice. He distinguishes between enumerative studies and analytic studies. An enumerative study has for its aim an estimate of the number of units of a frame that belong to a specified class. An analytic study has for its aim a basis for action on the cause-system or the process, in order to improve product of the future. Techniques and methods of inference that are applicable to enumerative studies lead to faulty design and faulty inference for analytic problems. It is possible, in an enumerative problem, to reduce errors of sampling to any specified level. In contrast, in an analytic problem, it is impossible to compute the risk of making a wrong decision. A number of examples clarify the issues.
This seven-page article appeared in The American Statistician, Vol.29, No. 4, 1975, pp. 146-152.
In this paper Dr. Deming continues with some of the practical problems of sampling that are not in the books. He writes on statistics as a basis for action; including the frame, the universe, and environmental conditions; the steps in the design of a study; statistical controls for detection of nonsampling errors; differences between investigators; replicated designs for simplicity in computation of variances; enumerative and analytic studies, contrasted; two kinds of error in an enumerative problem; two kinds of error in an analytic problem; limitations of statistical inference; criticisms of teaching; use of judgment samples; sampling for a rare characteristic; interpreting standards.
This 24 page article was prepared for presentation and criticism for the Princeton Conference on Applied Statistics, sponsored by the Metropolitan Section of the ASQC and the Princeton Biopharmaceutical Subsection of the American Statistical Association. Revised for the National Conference of the Statistical Reporting Service, Washington, 18 February 1975.
Dr. Deming wrote this paper to present a number of principles of training and administration that upset generally accepted conventions, and to point out to management that most of the trouble with faulty product, recalls, high cost of production and service, is chargeable to the system and hence to management. Effort to improve the performance of workers will be a disappointment until the handicap of the system is reduced. Dr. Deming includes sections on Road-Blocks to Quality in America, Loss From Variation, Two Sources of Variation, Thumb-Nail Sketch of the State of Statistical Control, Familiar Consequences of Faults of the System, Some New Principles in Administration, and four examples.
This 15 page article appeared in Interfaces, Vol.5, No. 4 in August, 1975.
Dr. Deming wrote this paper to exhibit some of the consequences of failing to distinguish between the enumerative and analytic uses of data. The paper includes definitions of the enumerative and analytic uses of data, gives examples of each, and makes special reference to the statistical control of quality. He then details the sampling variances for the two types of distributions, and discusses the four possible different variances, acceptance sampling, and consumer's risk.
This 12 page paper was published in the Journal of the American Statistical Association, Vol. 48, 1953, pp 244-255.
Dr. Deming highlights six issues in this paper: (1) the need for quality, as exhibited by increasing deficit in trade in the U. S., (2) bad practices of management, (3) failure of management to accept responsibility for quality, (4) changes required in the teaching of statistics, (5) enumerative and analytic studies, and (6) differences between use of complete count in a study and a sample less than 100%, and who is responsible for making decisions of different types in such a study
This fifteen page paper was delivered at the meeting of the International Statistical Institute in Tokyo 8-11 September 1987.
In this paper, Dr. Deming aimed to set forth a few points on the teaching of statistical principles and techniques to people in industry at four levels: Management, Statistical Administration, Research, and Workers in the Manufacturing Plant. Each of these four groups has an important, but different, role in industry. Management needs to know the principles of statistics and to have an appreciation for the uses of statistical techniques. The statistical administrator must have a keen perception of statistical problems, some knowledge of the techniques themselves; an intimate knowledge of the problems that confront the management and the ability to put statistical techniques to work to solve these problems. Research people need to possess penetrating knowledge, skill in the techniques, and originality to develop new theory where required. Production people need to understand simple techniques like control charts and sampling. Dr. Deming elaborated on each of these recommendations in this work.
This 24 page article appeared in the Bulletin of the International Statistical Institute, Tome xxxiv - 2nd Part, 1954. He had presented this work at a conference of the Institute in Rome in September 1953.
Dr. Deming's aim in this article is to try to answer some of the questions that statisticians and others have been asking for years about judgment-samples: (1) What type of problem requires, for best efficiency, use of a judgment-sample of blocks, clinics, hospitals, machines? (2) What is the best allocation of the sample? Sections focus on: enumerative and analytic studies, contrasted; two kinds of error in an enumerative problem; what we need in an analytic problem; two kinds of error in an analytic problem; limitations of statistical inference; use of judgment-samples; importance of design of experiment.
This six-page article appeared in Statistical Applications of Research (Union of Japanese Scientists and Engineers, Tokyo) vol. 23, 1976, pp 25-31. It was delivered at the Institute of Statistical Mathematics, Tokyo, Nov. 1975.
The purpose of this article is to bring to the attention of experts and practitioners some neglected points concerning the way in which statistical methods are being used in many fields. The ultimate aim is to loosen up some of the roadblocks that seem to stand in the way of proper use of modern statistical techniques. Dr. Deming focused on the purpose of sampling, sampling as cooperative work between the subject-matter expert and the statistician, the statement that "our problems are different", some everyday uses of sampling, the eight parts of a sampling plan, and some misconceptions about sampling.
This 14-page article appeared in the Interstate Commerce Commission Practitioners' Journal, Vol. XXXI, Nov. 1963.
The theme of this paper is the source of power of statistical theory, how to use it in industry and in research, and more especially, how not to misuse it. Dr. Deming discusses the value-judgments in the management of business, and the perception of the social, political, and economic problems that a business encounters in the achievement of these aims. He talks about judicious choice of problems and clever separation of the responsibilities for solving them.
This paper is an address delivered by Dr. Deming upon receipt of the Shewhart Medal at the 10th annual meeting of the American Society for Quality Control, Montreal, 7 June 1956. It was published as a three-page article in Industrial Quality Control, Vol XIII, No. 1, July 1956.
This study is part of the work of the U.S.-U.K. Cross-National Study of Diagnosis of the Mental Disorders. Reexamination of hospital diagnoses made in 1932-1941 and 1947-1957 by American-trained psychiatrists and a British-trained psychiatrist reveal differences that the authors attribute to inconsistent diagnostic criteria over time. The authors believe there may have been a decrease in the number of "hard-core" schizophrenics while the number of "ambiguous" schizophrenics increased, inviting more diagnoses of schizophrenia by those with a broader concept of schizophrenia.
This six-page article was co-authored by Judith B. Kuriansky, W. Edwards Deming, and Barry J. Gurland, and presented at the 126th annual meeting of the American Psychiatric Association, Honolulu, Hawaii, May 7-11, 1973. It was published in the American Journal of Psychiatry 131:4, April 1974. Ms. Kuriansky and Dr. Gurland were with Biometrics Research, New York State Department of Mental Hygiene and the Department of Psychiatry, Columbia University.
The aim of this paper by Dr. Deming is to compare several plans of sampling that often appear to be equal, but which may give widely different degrees of precision when put into use. After describing three plans to estimate a total population, he covers estimates of a ratio, and the effect of blanks in the frame. An appendix gives an illustration of a frame with two units.
This eighteen-page article appeared in Contributions to Applied Statistics. dedicated to Professor Arthur Linder, Edited by Walter Joh. Ziegler, 1976 Birkhauser Verlag, Basel und Stuttgart.
In this paper Dr. Deming aimed to explain some points about statistical design and inference that had not yet found their way into textbooks. Some fifteen years later, these concepts are still not in most textbooks. The article is not about statistical techniques, but is about how to understand the meaning of statistical results. Sections include the purpose of the study (enumerative and analytic), the frame of study and when it is satisfactory, the equal complete coverage, nonsampling errors, three uncertainties that may exist, the meaning of standard error, limitations of statistical inference, and need for operational definitions.
This 16 page article appeared as Chapter 20 in the book The Behavior of Psychiatric Patients, Quantitative Techniques for Evaluation, edited by Burdock, Sudilovsky and Gershon, published in 1982 by Marcel Dekker, Inc., NY.
While a patient, Dr. Deming observed that the administrator of a hospital knows a lot about what happens in the hospital. So does a head nurse. The physicians know much about the hospital that no one else knows. A patient in the hospital sees what no one else sees. All these different points of view, were they known, might be helpful to the management of a medical care system. A hospital is an important component in a system of medical care, and needs all of these inputs. In this article Dr. Deming points out the need for the view of medical care as a system. He then goes on in his inimitable style to illustrate the point with his (the patient's) view of one particular hospital experience.
This four page article appeared in the Journal of the Society for Health Systems, Volume 2, Number 1, Spring 1990.
Dr. Deming states that "The purpose of this (paper) is to help people that do research, and people that pay for research, to get more for their money." He discusses the answers to four questions: (1) Does a figure convey information? (2) What is the aim of good design? (3) Is good statistical work costly? (4) Can you distinguish a probability sample from something else? He then focuses on three types of uncertainty in data, variances between interviewers, and use of the standard error. In conclusion he debunks six fables about surveys.
These ten pages are an address delivered by Dr. Deming to the Market Research Council in New York on the 17th September 1971.
Dr. Deming wrote this paper to fill in some of the missing links in the use of statistical methods, with special reference to responsibilities at the management level for effective mobilization of statistical knowledge and skills. He describes statistical control of quality as a system, not a bag of techniques, focuses on statistical methods in improvement of operations, and describes 24 uses of statistical methods in various stages of production.
This 22 page paper was delivered by Dr. Deming at the All India Conference on Quality Control, New Delhi, 17 March 1971. It is reprinted from the Proceedings, pages 98-119.
Most frames for censuses, complete testing, or sampling, are already stratified into natural geographic zones. Material, as it comes to us in the frame, is never thoroughly mixed. This natural stratification is automatic and costs nothing. It sometimes pays to rearrange the sampling units into more homogenous groups called strata. We sometimes rearrange all the sampling units in the frame before we draw the sample, and sometimes we rearrange only the sampling units in the sample, depending on costs. In this technical paper, Dr. Deming describes nine possible procedures of stratified sampling, indicates when each is most appropriate and compares the gains in precision with the costs for each plan.
This 23 page article appeared in Estadistica, Journal of the Inter-American Statistical Institute, December 1959, pp 716-738.
Dr. Deming wrote this article to present some theory by which to measure the influence of the inspector, and to show how sensitive the results of inspection are to this influence. The theory indicates what records to keep in order to improve performance of inspectors and the supervision of inspection, whether inspection be carried out by visual inspection, or by use of instruments, or by automatic recording devices. Dr. Deming and Dr. Morris H. Hansen then give an example of use of this theory to estimate the overall physical condition of gas pipe buried in the ground.
This twelve-page article appeared in Statistica Neerlandica, Vol. 26/3, 1972 (in English).
The purpose of this paper is to present, from Dr. Deming's point of view, why SQC has been so effective in Japan. He concentrates on the distinction between special causes and common causes, which appeared to be especially important. He shows that statistical techniques enable us not only to detect the existence of conditions that need correction; they also determine which administrative level is responsible to identify and correct the cause of trouble. Dr. Deming defines SQC and discusses some responsibilities of management, the dynamic nature of statistical control, and the power and limitations of statistical theory.
This four-page article appeared in ICQC' 69-TOKYO.
This article is in the form of a letter to a manager who has requested help from Dr. Deming on some problems with production, high costs and variable quality. Dr. Deming states that failure of management to accept and act on their responsibilities in quality control is one cause of trouble. He continues with 34 observations of how this plays out in the organization.
This three-page article appeared in Quality Progress, Vol 5, no 7 in July 1972. It was written in response to a Viewpoints column on process capability.
Dr. Deming wrote this article to explain some of the problems in the design and interpretation of a study whose aim is to evaluate the effectiveness of some treatment or plan; also to point out some of the difficulties of studying by retrospect the cause of success or failure, or the cause of a disease or of a specific alleged cure therefor. Emphasis is on ways to improve the reliability of evaluation by understanding and avoiding possible misuses of statistical techniques in evaluation. Sections focus on the issue of evaluation, need for care in definitions of terms, four requirements for an effective system of evaluation, limitations of statistical inference, enumerative studies contrasted with analytic studies, possible mistakes in an enumerative study, possible mistakes in an analytic study, use of judgment samples, statistical tests of hypotheses, and the retrospective method.
This sixteen-page article appeared as chapter 4 of the book, Handbook of Evaluation Research, Vol. 1, edited by Elmer L. Struening and Marcia Guttentag (Sage Publications, 1975).
In this article Dr. Deming focuses on the crisis of Western Industry and some of the actions that management can take to overcome the crisis. A discussion of how the declining market exposes weaknesses is followed by a list of some of the forces that feed the decline. Dr. Deming includes remarks on evaluation of performance, use of visible figures, and other obstacles. Modern principles of leadership are put forth, along with a condensation of the 14 points for management.
This six-page article is in Volume 7 of Handbook of Statistics, edited by P. R. Krishnaiah and C. R. Rao (Elsevier Science Publishers B.V. (1988), pp 1-6.
The purpose of Dr. Deming's paper is to point out to statisticians the necessity that they be the logicians and architects of a survey or experiment. Statisticians need to classify the responsibilities in the planning, execution, and interpretation of the results. The paper also points out some principles for guidance. Five sections cover (1) The Universe, the Frame, and the Equal Complete Coverage; (2) Operational Definition of the Expected Value and of the Standard Error; (3) Classification of Uncertainties, and Responsibilities for Reducing Them; (4) Operational Definitions of the Bias and of the Accuracy of a Technique; and (5) Examples of Statistical Reports.
This 19 page article was published in the Bulletin of the International Statistical Institute, Vol. XXXVIII, Part IV, Tokyo 1961, 365-383.
Dr. Deming wrote this article to offer some observations on the causes of success in Japan, from the viewpoint of the statistical control of quality. He believed that appreciation of what happened in Japan might lead to successful programs in other parts of the world. This paper describes the nine reasons Dr. Deming credited for the success and speed of application of the statistical control of quality in Japan. He also translates into action the definition of statistical control, discusses his lectures to top management, and writes about the power and limitations of statistical techniques.
This five-page article appeared in Industrial Quality Control, Volume 24, No. 2, in August 1967 and was dedicated to his friend and colleague, Walter A. Shewhart, who had recently died.